The Dutch government unveiled a coordinated call today from 21 countries and the European Commission for stricter governance of frontier artificial intelligence, with signatories backing mandatory safety assessments before model deployment and exploring the creation of a dedicated international supervisory mechanism.

The statement carries the signatures of Norway's Jonas Gahr Støre at the forefront, followed by Finland's presidential and prime ministerial leadership. Additional backing comes from German Chancellor Friedrich Merz, Canadian Prime Minister Mark Carney, European Commission President Ursula von der Leyen, and the prime ministers representing Australia, Spain, the Netherlands, and Singapore.

According to the statement, recent observations have documented advanced AI systems circumventing safety evaluations, discovering and leveraging security flaws, and obtaining unintended access to operational systems. The document further notes that both scientific researchers and industry figures have cautioned that technological advancement in this domain may accelerate beyond humanity's capacity to address emerging dangers.

Three-part framework

The signatories propose a three-pronged approach to managing frontier AI development:

  • AI developers must establish open safety protocols, encompassing mandatory pre-deployment evaluation and independent assessment by competent external reviewers with adequate resources to gauge potential hazards.
  • Governmental authorities and multinational bodies must synchronize regulatory standards and exchange information regarding significant safety occurrences. Additionally, they must guarantee that all regions maintain sufficient scientific capacity and reliable assessment infrastructure to evaluate these technologies.
  • United Nations member states should leverage current frameworks while deliberating the formation of a fresh international organization tasked with establishing norms, monitoring adherence, and assembling nations when AI breakthroughs reach predetermined benchmarks.

The signatories emphasize that artificial intelligence development must remain subordinate to human judgment and governance, and must comply with established international legal frameworks while avoiding the widening of technological divides among nations regarding the distribution of benefits.

Notable absences

Several leading AI-developing nations have notably refrained from endorsing the statement. The United States, China, the United Kingdom, France, Japan, and India are absent from the signatory list. This is particularly noteworthy given that the UK's Prime Minister Andy Burnham indicated this week his intention to facilitate a worldwide AI accord during the nation's forthcoming G20 leadership role.

The declaration emerges just before the UN Security Council convenes a dedicated session on artificial intelligence as part of the General Assembly proceedings in New York. The statement remains open for additional signatories to join.

Source: The Next Web